plingbang

I had an idea for a project I called "LLM 2007" - to produce a CD that would include an LLM that would run on a PC of that era. I wanted to make it look as if it really was a mid-2000s product. That would make it seem like an artifact that came from alternate reality.

For that, I thought of making a Windows 2000 style GUI using pure WinAPI, backporting llama.cpp and bundling Qwen3-0.6B. Never started this though.

This feature has been was available on GNOME since forever. I'm still waiting for Microsoft to steal another nice thing: MRU (Most Recently Used) keyboard layout switching. If you have two often used keyboard layouts A and B, and less used layouts C, D, E, then pressing the switch shortcut once will switch between A and B without the need to cycle through C, D, E.

I've got an identical output with the prompt "do not lock the door".

I wonder if the sound could be fully synthesized instead of the current 2.9s recording. That's not an easy task, but perhaps the result would sound more realistic.

> Teen-by-default settings to roll out globally for all Discord users

Does it mean that even people who reside outside jurisdictions touched by the age verification craze will have to deal with all this?

> use facial age estimation

Surely a kid won't be able to ask someone else to pass the check for them. But let's talk about false positives. If the estimator falsely declares someone an adult, is Discord legally liable?

> submit a form of identification

If you have a picture of an ID document, can you verify that it's real? You'd have to ask the government for that. And at least in one country there is no process for that.

> On-device processing

Oh, a client-side check. Must be secure.

I have a recording of me trying to pass the captcha for straight 5 minutes and giving up. To be fair, this has only happened once.

What is the purpose of such loop? Bots can simply switch to another residential proxy when the captcha success rate gets low. For normal humans, it is literally "computer says no".

citation needed

A very close idea was jokingly suggested in The Birth & Death of JavaScript[1] at 18:07 (rewind to 14:14 for more context)

[1]: https://www.destroyallsoftware.com/talks/the-birth-and-death...

Fighting with the OS is futile. The OS is always in control and apps can only ask it nicely to do things.

Microsoft can simply change Recall to capture DRM-marked content too. And to avoid copyright issues, it will store some kind of visual summary (or whaterer the neural network can use) instead of plain screenshots like it is doing now.

I found a wall of shame that includes manufacturers that make it hard or impossible to unlock the bootloader[1]. It was mentioned in a forum post[2].

[1]: https://github.com/melontini/bootloader-unlock-wall-of-shame

[2]: https://xdaforums.com/t/future-of-unlocking-bootloaders-in-a...

> It's tricky, though. What else can you do?

I had an idea about amost-privacy-preserving system by involving government ID and blind signatures:

1. The service passes a random string to the user. 2. The user authenticates to their government and asks the government to sign it. 3. The government applies a blind signature which basically says "this user/citizen hasn't registered an account in the last 60 minutes". 4. The government records the timestamp. 5. The user passes the signature back to the service.

Upsides:

* Bypassing this would be orders of magnitude more expensive than phone numbers. * Almost private

Downsides:

* Won't happen. Remote HW attestation is likely to win :( * The service knows your citizenship * The gov knows when and how often you register. * Any gov can always bypass the limits for themselves.

I think it may be also possible to extend it so that the government attests that you have only one account on the service but without being able to find which account is yours.

Some time ago there was a website that showed you a random YouTube video. Like truly random. The biggest discovery to me was that a typical video has 0-1 views, nearly always <10. I bet most people don't realize this is how YouTube actually looks like. And I guess it's also a good small reminder to all people trying to become famous on social media.

I believe the website tried to find videos with least bias possible by doing some clever searches using YouTube API (so not just videos titled IMGXXXX). Maybe it was trying to do partial matches on video ID.

Looks usable and useful unlike my joke version:

https://plingbang.github.io/morsel/

That's probably rare but I had a no name TV which just let me just enable adb over network with full root access. IIRC I had to install an app that can launch arbitrary activities so I could access the buildin Android settings menu instead of the crippled TV settings UI.

For a case when file sharing is intended between individuals or small groups there's an easy solution:

Anyone who got the link should be able to delete the file.

This should deter one from using the file sharing tool as free hosting for possibly bad content. One can also build a bot that deletes every file found on public internet.

Can you provide some context? Do you mean MS employees are prohibited from doing open source even outside work hours?

Awesome but seem to fail on some numbers e.g.

https://api.isfizzbuzz.xyz/api/15000000000000000000000000000...

BTW, there are neat divisibility rules which can give you the answer in practically linear time when the number is in decimal: https://en.wikipedia.org/wiki/Divisibility_rule

That's a good point. Moreover, someone built[1] an SSH server that prints your name when you connect (because GitHub publishes SSH public keys of every user):

    ssh whoami.filippo.io

[1]: https://words.filippo.io/ssh-whoami-filippo-io/

> so users can see in real-time what code was run and when

I'm really curious how this was presented to the user. A table with timestamp, filename, line number, and line contents? Or something more advanced?

I understand they "cannot test every possible browser" and that "users may get subpar experience".

I don't understand why there isn't "continue at your own risk" button. Maybe with a scary warning. It's kind of stupid that I have to spoof UA for a website to let me in. And in most cases, everything just works fine.

Maybe one day I'll create a website to inform about the issue.

Can you name a single Android phone or tablet model that is <10 y.o. for which you can access personal data with this method?

Why not just an LLM-based interpreter that direclty executes a PDF spec plus edits received by email? No need to recomplile and restart the app. A DB is also not required - the LLM will naturally remember all user requests and figure out the current state. (We'll solve the limitations of context later)

And what makes it worse in phpBB and other forum engines is the waste of vertical space. 4 messages on your screen seems to be the maximum despite using smaller fonts than web average nowadays.

> Or you can end up like the people who lost their data [1]

I don't see how encryption at rest could've changed the outcome.

In the article, the cloud provider, which has full control over the VMs, was compromised. The VMs were hosting various Bitcoin services, which needed continuous wallet access for operation. So, I'd say there was no data at rest to be secured. The attackers could theoretically patch the application to make malicious transactions or just extract the wallet from RAM.

Also, the article suggests that the attackers were getting inside the running VMs rather than accessing VM storage directly.

I remember MS Teams opening various files within itself instead of launching an appropriate program. I wonder what they're using for rendering PDFs.

I used gnome-terminal for years and at the times when I had to switch to a virtual console, I often had a feeling that the latter was more responsive.

But that could be the placebo effect due to higher cursor blinking and key repeat rates. My monitor is 60 Hz.

Sounds like Selection sort[1]. It was probably the first sorting algorithm I ever implemented. I don't know why bubble sort is usually taught first. Selection sort seems more obvious. Perhaps the implementation is slightly harder due to the need to keep track of two indexes.

1. https://en.wikipedia.org/wiki/Selection_sort

Combined with IP source address spoofing, it would probably help greatly with DDoS amplification attacks while only saving up <50% packets for good users.

If you care about time rather than packet count, you can send packets with all reasonable TTL values at once.

Interestingly, Firefox records all your visits to a particular page, not just the latest visit. You could've changed the history sort order to "first visit" so that the list doesn't change when you revisit the page but that option simply isn't there in the UI.